HomeAI models
Perplexity: Sonar Pro
Perplexity
•
Sonar Pro
•
Released 
January 2025

Perplexity
Sonar Pro

Search-grounded model with real-time web retrieval and citations, built for factual answers over live sources rather than trained knowledge.

Modality:
Text
Image
model ID
perplexity/sonar-pro
Start for free
arrow black icon
View API Docs
arrow white icon

Output Speed *

N/A
tok/s

Intelligence Index *

7.6
/ 100

Context Window *

200000
tokens

Input price

18
Anytoken

Output price

90
Anytoken
Sonar Pro: Factual Answers Grounded in Live Web Search Sonar Pro is Perplexity's search-augmented answer model. Rather than answering from stored training data alone, it retrieves and synthesizes multiple live web searches and returns cited responses. It sits above the base Sonar tier in Perplexity's lineup, adding a larger context window, deeper multi-step query handling, and roughly double the citations per search. The workload that benefits most is real-time, source-grounded question answering—research assistants, competitive monitoring, and factual lookups where traceability and current information matter more than raw generative reasoning. Access Sonar Pro through the AnyAPI.ai API

Performance

Where Sonar Pro Earns Its Place: Grounded Factuality

Sonar Pro is optimized for factual, source-grounded answers rather than open-ended generation. Because it answers from live web retrieval instead of stored parameters, it leads factuality tests for search-enabled question answering: Perplexity reports an F-score of 0.858 on SimpleQA versus 0.773 for base Sonar. For developers, this means fewer fabricated facts on short, fact-seeking queries and every answer arriving with traceable citations. In production, that shifts verification effort from manual fact-checking toward reviewing cited sources—valuable for research assistants and monitoring tools, but it also makes the model dependent on retrieval quality rather than reasoning depth.

Benchmarks

Sonar Pro on Factuality and Independent Reasoning Benchmarks

On SimpleQA, the standard short-form factuality benchmark, Perplexity reports Sonar Pro at an F-score of 0.858, ahead of base Sonar's 0.773—an advantage attributed to real-time retrieval rather than parametric knowledge. On general reasoning, independent aggregators place Sonar Pro more modestly: reported figures include roughly 75.5% on MMLU-Pro and 57.8% on GPQA Diamond, with weaker coding results near 27.5% on LiveCodeBench. Independent analyses have also flagged citation-attribution errors as a distinct risk, where real URLs are paired with inaccurate claims. Treat Sonar Pro as a factuality-and-retrieval specialist, not a general reasoning or coding leader.

Output Speed

*
N/A
tok/s

Intelligence Index

*
7.6
/ 100

MMLU *

Broad world knowledge and problem-solving
76
%

GPQA *

PhD-level scientific reasoning across physics, biology, chemistry.
58
%

HLE *

Adherence to multi-step structured instructions.
7
%

LiveCodeBench *

Tool-calling reliability in long agentic loops.
28
%

Technical Specifications

What the model supports

Sonar Pro accepts text and image input and returns text, with a 200,000-token context window and up to 8,000 output tokens per response. The two specifications with the largest production impact are the modest output ceiling and the built-in, always-on web search: the 8K completion limit constrains long report generation, while integrated retrieval means answers reflect current web content and arrive with citations. Structured JSON-schema outputs are available on select tiers. Note that capabilities like function calling vary by endpoint and hosting provider, so verify tool support against the specific route you use.
Verified Specifications — 
Sonar Pro
*
Input modalities
Text
Image
output modalities
Text
Context window
200000
 tokens
Maximum output tokens
8000
Reasoning
No
Knowledge cutoff
January 2025
Pricing (standard)
18
 AnyTokens in
 / 
90
 AnyTokens out

Limitations & Trade-offs

Where Sonar Pro falls short

1
Short output ceiling. Sonar Pro caps completions at 8,000 tokens, which is small for long-form report generation. Workloads that need multi-page synthesized documents in a single call—exhaustive research reports or long structured extracts—will hit this limit and require chunking or a research-focused alternative. For question answering and summaries the ceiling is rarely a problem, but plan output length deliberately when the deliverable is a long document.
2
Citation-attribution risk. Because Sonar Pro presents real source URLs, factual errors are harder to catch: independent analyses have flagged cases where genuine links are paired with claims those sources do not fully support. This makes hallucinations less obvious than in models without citations. For high-stakes use—legal, medical, financial—treat citations as leads to verify, not as proof, and keep human review in the loop.
3
Layered per-request pricing. Beyond token costs, Sonar Pro adds a search request fee that scales with search context size (Low, Medium, High). On short, high-volume queries this fee can exceed the token cost itself. For cost-sensitive, high-throughput deployments, set search context size per query rather than globally, and consider base Sonar where retrieval depth is not needed.
4
Not a reasoning or coding specialist. Sonar Pro's strength is retrieval-grounded factuality, not multi-step logic or code generation—independent figures show comparatively weak coding results. For chain-of-thought analysis choose Sonar Reasoning Pro, and for heavy coding tasks a general frontier model is a better fit. Use Sonar Pro where the value is fresh, cited information, not derived reasoning.

Best-Fit Workloads

Where this model earns its place

01

Real-time research assistants

‍
Sonar Pro's live retrieval plus roughly double the citations of base Sonar makes it well suited to research copilots that must answer from current sources with traceability. Its 200,000-token context supports longer follow-up conversations and multi-source synthesis. Teams building sales or market research tools benefit from grounded, cited answers; just verify citations for high-stakes claims.

02

Competitive and regulatory monitoring

‍
For tracking competitor launches, security advisories, or regulatory changes, Sonar Pro combines recency with source attribution, letting you wire domain filtering and structured JSON output into alerting pipelines. The always-on web search removes the need to stitch a separate search tool onto a general LLM, simplifying the architecture for continuous monitoring jobs.

03

Factual question answering

‍
Its leading SimpleQA factuality score (0.858 F-score, Perplexity-reported) makes Sonar Pro a strong fit for fact-seeking Q&A features where correctness on short queries matters—product help, knowledge lookups, and API/documentation questions. Answers arrive with citations, shifting verification from manual fact-checking to source review. Keep outputs within the 8,000-token limit for these concise queries.

04

Grounded RAG augmentation

‍
Where an internal RAG stack lacks fresh public data, Sonar Pro can supply live, cited web context to complement private retrieval. It effectively acts as a managed web-retrieval layer with attribution, reducing the need to build and maintain a separate crawling and ranking pipeline. Factor the per-request search fee into high-volume RAG designs.

Pricing in anytokens via AnyAPI
Input
18
₳
Output
90
₳
Cache write
—
₳
Cache read
—
₳
Get API access
arrow black icon
Read Docs

on this page

PerformanceBenchmarksTechnical SpecificationsLimitations & Trade-offsBest-Fit WorkloadsIntegrationFAQ

Integration

Access Sonar Pro via AnyAPI.ai

Access Sonar Pro through AnyAPI.ai using a unified API built for multi-model AI applications. Integrate Sonar Pro without maintaining a separate provider-specific connection, and keep the flexibility to test, switch, or combine models as your application requirements evolve.

01

One API integration

Access Sonar Pro and other AI models through the same API workflow instead of maintaining separate integrations for every provider.

02

Easy model switching

Test Sonar Pro against alternative models or switch models as your performance, capability, or cost requirements change without rebuilding your application around another provider API.

03

Flexible for production

Use Sonar Pro from experimentation through production while keeping your AI stack flexible as workloads, traffic, and model requirements evolve.

04

Multi-model applications

Use Sonar Pro for the workloads where it performs best and combine it with other models for tasks that require different capabilities, performance, or efficiency.

Start free
arrow black icon
Read Docs

Frequently Asked Questions

Answers to common questions about integrating and using this AI model via AnyAPI.ai

Sonar Pro is Perplexity's search-augmented answer model that responds to queries using live web retrieval rather than only trained knowledge, returning answers with citations. It sits above the base Sonar tier, adding a 200,000-token context window, multi-step query handling, and roughly double the citations per search, making it suited to real-time, source-grounded question answering.

Sonar Pro has a 200,000-token context window and supports up to 8,000 completion tokens per response. The large context suits longer, multi-source searches and follow-up conversations, but the 8,000-token output ceiling limits long-form report generation in a single call, so plan chunking for lengthy documents.

Perplexity reports Sonar Pro at an F-score of 0.858 on SimpleQA, the standard short-form factuality benchmark, ahead of base Sonar's 0.773. The advantage comes from answering via real-time web retrieval rather than stored data. Note that independent analyses have flagged citation-attribution errors, so cited sources should be verified for high-stakes use.

Sonar Pro supports structured outputs via JSON schema on select tiers, and web search is built in and always active. Function/tool calling availability depends on the specific endpoint and hosting route, and some routes do not accept tools, so verify tool support against the exact endpoint you use.

Use Sonar Pro when queries need multi-source synthesis, richer citations, longer context, or higher factual precision—research assistants, competitive monitoring, and compliance-sensitive lookups. Use base Sonar for simple, high-volume, latency- and cost-sensitive lookups, since it is cheaper on both tokens and per-request search fees.

* Benchmark data source: Artificial Analysis artificialanalysis.ai